variable pay systems: collective bargaining and wage ... · variable pay systems: collective...
TRANSCRIPT
1
Variable pay systems: collective bargaining and wage inequality in six different European countries
Rosa Garcia-Hernández
e-mail: [email protected]
(UAB, Autonomous University of Barcelona, Applied Economics Department and UB, University of Barcelona, Economic Policy Department)
1 Introduction1
The aim of this paper is to analyze the impact of Variable Pay Systems in wage
determination and in wage inequality, in the case of six European countries:
Finland, Spain, Portugal, France, Romania and Poland. We use data from SES
(Structure of Earning Survey) to carry out our econometric analysis, which
offers a cross-sectional dataset and includes matched employer-employee
microdata.
According to literature, Variable Pay or Pay for performance, supposes
additional components to regular wages in order to improve productivity or
motivation workers. But our first conclusions indicate that, in some cases,
these remuneration systems could be a variable which worsens the wage
distribution and which contributes to grow wage inequality.
1.2 Motivation
European Company Survey (2013) (Eurofound, 2015) showed that the 63% of
European analyzed establishments used some kind of Variable pay systems.
These schemes of variable remuneration have had a growing importance over
last years, as different papers show us (Pendleton A., Whitfield K. Bryson A.,
2009).
1This Research Project is part of the Department of Economic Policy and World Economic Structure research project of UB called "Analysis and evaluation of public policies" ECO2012-38004 from Spanish Ministry of Economy and Finance. Eurostat-Research Proposal 53/2015-SES)
2
Table 1. Percentage establisments with Variable Pay Systems by classification
Payment results
PRP-Individual
PRP-Team
Profit-sharing
Share-ownership
LT 72.2 CZ 74.3 EE 49.1 SI 55.4 FI 12.5
CZ 57.9 SI 72.5 ME 48.3 ME 53.7 LT 12.8
EE 56.9 ME 70.2 LT 47.6 LT 53.1 LU 12
AT 53 LT 67.2 SI 47.9 FI 50.8 MK 9.3
ME 51.9 MK 56.4 BG 41.3 AT 45.8 SE 9.2
SK 50 AT 56.4 SK 40.1 EE 41.8 FR 8.1
MT 46.3 SK 55.2 PL 39.7 FR 40.9 UK 8.6
FI 45.6 PL 54.6 LV 31.8 SE 37.6 TR 7.3
IS 43.1 DK 53.4 RO 29.2 DK 35.2 EE 7.9
RO 40.2 LV 48.4 TR 28.4 NL 34.2 AT 6.6
SI 40.2 NL 47.8 FR 26.3 BG 33.9 CY 6.0
PL 39.1 RO 45.3 PT 25.3 IS 31.7 IE 6.5
FR 38.5 DE 44.2 EU 25.2 EU 30.1 ME 5.6
NL 38.7 FI 44.5 UK 25.1 LU 28.8 EU 5.2
LU 37.7 IS 43.5 NL 24.2 UK 26.1 BE 4.9
DK 36.4 MT 43.4 ES 23.2 ES 25.1 ES 4.7
UK 36 EU 43 IE 22.5 TR 24.0 BG 4.5
HR 35.2 UK 41 EL 20.6 IE 23.5 CZ 4.4
BG 34.5 HR 40.3 HR 20.7 LV 22.6 PT 3.6
EU 34 BG 39.7 DE 18.3 CY 21.9 DE 3.0
ES 33.9 EL 39.3 BE 17.8 BE 19.4 RO 2.8
DE 31.2 ES 35.1 IT 17.7 EL 17.3 EL 2.4
EL 31.8 IT 35.2 IS 15.5 HU 16.6 LV 1.7
IT 18.4 BE 31.6 HU 15.3 MT 13.4 MT 0.2 Source: Own elaboration from ECS 2013 (Eurofound, 2015)
Payment results
PRP-Individual
PRP-Team Profit-sharing Share-ownership
Intermediate + Rather Decentralized CB Intermediate CB
Finland 45,6 Poland 54,6 Poland 39,7 Finland 50,8 Finland 12,5
Romania 40,2 Romania 45,3 Finland 33,6 France 40,9 France 8,1
Poland 39,1 Finland 44,5 Romania 29,2
Intermediate + Rather Centralized CB
Intermediate+Rather Centralized+Rather Decentral.CB
France 38,5 EU 43 France 26,3 Poland 34,3 EU 5,2
EU 34 France 39,7 Portugal 25,3 Romania 31,7 Spain 4,7
Spain 33,9 Spain 35,1 EU 25,2 EU 30,1 Poland 3,3
Portugal 27,1 Portugal 34,7 Spain 23,2 Spain 25,1 Portugal 3,6
Portugal 21,7 Romania 2,8
Total 63%
34%
43%
25%
30%
5%
3
Payment by results, Performance Related Pay, Profit Sharing and Share
ownership are forms of variable remuneration which are important for
different reasons. On the one hand, because their growing importance in the
collective agreements. On the other hand, because, although they could be
considered as a type of labor demand factor, connected for example with the
evolution of the firm objectives, some of them are connected with the
evolution of the individual features and productivity. Some literature justifies
the introduction of Variable Pay Systems with the improvement in
productivity, because of their connection with motivation workers. But the
purpose of this paper is to analyze if this supposed effect on productivity goes
with an increase in wage inequality in some European countries. Following
some papers (2010, Bryson, Freeman, Lucifora et al) in which in the case of
Italian metal engineering firms the introduction of pay for performance
increase wage inequality by around 3%.
So our research question is: using variable pay or pay for performance implies
more or less wage inequality?
1.3 Literature review and contribution
1.3.1 Contribution
There are some literature that analyzes which factors are most relevant in wage
determination like Palacio and Simón, 2002. And other literature which
focuses on the analysis of the wage inequality through, for example,
decomposition of individual variance (Palacio and Simón, 2004), or Fields
decomposition (Simón, 2009). Lemieux, 2007 introduce Variable Pay System
into the analysis of wage inequality, taking data from Panel Study of Income
Dinamics. The main contribution of this chapter is to introduce Variable Pay
System to all theses previous analysis, using three last waves of SES.
1.3.2 Wage determination
Wages determination is a recurrent issue in Labour Economics (Katz y Autor,
1999). From neoclassical competitive model of labor market, wages are
determined by worker productivity level (marginal productivity labor). This
statement would have connection with the neoclassical theory of distribution
which tells us that the remuneration of production factors is equal to its
marginal productivity. So, in the determination of wage levels labor supply
factors should be predominant, while labor demand factors would play a
4
minor role or no role (Reder, 1962). Moreover, following this model, workers
with similar productivity levels should receive the same wages, regardless of
where they work. (Palacio y Simón, 2004).
This approach is connected with human capital theory (Becker, G, 1964)
(Schultz T., 1960) (Mincer, J.1958)2 and, according to which, wages depend on
education level, experience or seniority, for example (labor supply factors).
Also, age or gender can be included.
We could find this individual point o view of the wage determination in
Industrial Relations field. Here, wages would be “pay for the person” = base
pay ex ante productivity + additional pay-worker's productivity (Lemieux et al,
2007) and we have to talk about an individual variable pay, connected with
individual level of productivity, results or individual performance (merit pay).
However, we can find a lot of examples about the difficulty of companies to
pay people according to their marginal production (Kerr S., 1975).3
And different empirical evidence shows that labor demand factors offer us a
better explanation of wage inequality in developed countries than labor supply
factors: in the case of inter-industry wage differences, Groshen (1991b) or
Jaumandreu (1994)4. These labor demand factors are connected with
establishment features like: size, ownership, market, industry, etc. So, wages
determination will be influenced by other components different from
individuals (Palacio J.I., Simón H., 2004).
On the one hand, wages are “attached to jobs”, because compensation is
determined by the characteristics of the job or workplace (Lemieux, T. et al.
2007), like type of workday or responsibility level. Here variable pay would be
a collective performance pay connected with collective productivity or
performance or with team work. On the other hand, wages would be
“attached to company or factory features”, like ownership, company size,
industry, market, industrial relation regulations (labor demand factors). In this
case we are in front of a collective performance pay, linked to company
productivity, company results or company performance (profit related pay or
bonuses, share ownership schemes).
2 See Laroche M, Mérette M., Ruggeri GC (1998-01) 3 See Lemieux, T, Bentley MacLeod W., Parent D. (2007) 4 See Palacio y Simón (2004)
5
So, theories and authors explain that the wage determination goes further
market supply factors and market demand factors from labor market.
Institutional aspects and social actors, like trade unions and employer
organizations, have to be taken into account, because wage determination is
done through collective bargaining but no in a competitive market. This is the
point of view of the Industrial Relations (Pendleton A. et al, 2009) analysis and
the some authors of Labor Economics (Pérez Trujillo M., Ruesga S. et al,
2009).
1.3.3. Wage Inequality
One dimension of inequality is income inequality and it refers to the inequality
of the distribution of individuals, household or some per capita measure of
income5. Lorenz Curve measures level of inequality and poverty and
divergence of a Lorenz Curve for perfect equality and the Lorenz Curve of a
given income Distribution is measured by some index of inequality like Gini
index (Heshmati, 2004).
Analyze and understand the wage inequality is a very important issue in labor
market because this is a key determinant of differences in living standards
(Simón H, EES2002) and of income distribution.
Some factors which could become an explanation of wage inequality could be:
individual characteristics (labor market supply point of view), workplace and
establishment characteristics (labor market demand point of view) and labour
market institutions. (Simón H, EES2002).
In order to assess the level of wage inequality we can find different tools. We
focus in variance of logarithms and Gini Index. Moreover, if we want to
determine which are the most important factors have influenced in wage
inequality, we have to use some kind of inequality decomposition.
There are several approaches to inequality decomposition. Traditional
methods or “a priori” methods (Cowell and Fiorio, 2010) include the
decomposition by income sources (Shorrocks, 1982) and by population
5 The 1990s signified a shift in research previously focused on economic growth, its
determinants. This change supposed focusing in issues of convergence or divergence of per
capita incomes to the long-term equalization or polarization of incomes across regions and
countries. (Heshmati, 2004).
6
subgroups (Shorrocks, 1984). First method estimates the contribution of
individual income components to the observed inequality and second method
measure inequality both within and between subgroups of the population
(Manna R. and Regoli A., 2012). Regression-based approaches go further
including any factor (economic, social, etc) that may drive the observed
inequality and can manage problems of endogeneity due to reverse causality.
Regression-based decomposition methodology was introduced in 1970’s
(Blinder, 1973; Oaxaca, 1973). Thirty years after, Fields (Fields, 2003a)
introduced a regression-based decomposition by income determinants through
the extension of the decomposition by income sources (Manna R. and Regoli
A., 2012).
Fields decomposition (Fields, 2003)6
We start from an income or wage (in our case) generating function
(1)
Where w denotes wages, Xj the j-th explanatory variable, bj its coefficient and ε
the error term. The Fields method estimates the share of the log-variance of
income that is attributable to the j-th explanatory factor (relative factor
inequality weight) as:
(2)
Where j is the coefficient of the j-th explanatory factor estimated from an
OLS multiple regression, σ2 (ln w) is the variance of the dependent variable
and cov (Xj, ln w) is the covariance between the j-th factor and the dependent
variable. (Manna R. and Regoli A., 2012)
In the Fields decomposition, Sj FIELDS represents the contribution of each
factor to total inequality.
1.3.4. Collective bargaining regimes
Due to absence of EU harmonisation and that country-specific institutions
continued to exist, we could group EU’s 27 members states into five clusters
industrial relations regimes. (ETUI, 2012) (European Commission, 2009):
6 See Simón, H (2009)
7
1)North European: Denmark, Finland and Sweden
2)Central-West European: Austria, Belgium Germany, Luxembourg,
Netherlands and Slovenia
3)South European: France, Greece, Italy, Portugal and Spain
4)Liberal-West European or Anglo-Saxon: Cyprus, Ireland, Malta and the UK
5)Central-East European: Bulgaria, the Czech Republic, Estonia, Hungary,
Latvia, Lithuania, Poland, Romania and Slovakia
These five groups are differenciated from each other in terms of some
elements. (ETUI, 2012)
-For the North, South and Central-West, multi-employer bargaining is
observed, between unions and employer associations (sector level bargaining).
For Anglo-Saxon and Central-East European states, single-employer
bargaining, between individuals employers and unions, is the norm (company-
level bargaining)
-In the North and Central-West European countries there are a relationship
between political actors and trade unions and employer associations. In Anglo-
saxon countries, social partners voice is not always reflected in policy
outcomes. In Southern Europe, the participation of social partners in policy
depends on governments’ willingness. In Central-Eastern Europe,
politicisation of social partners limits their influence in policy making.
-In Northern, Anglo-Saxon and Central-Western Europe, the state
involvement is not common. In Southern Europe, state influences collective
bargaining otucomes indirectly. (ETUI, 2012)
Looking to all the five different collective regimes, anyone could ask that
belonging to one or to other regime has any kind of influence in the
implementation of Variable Pay System.
8
1.4 Methodology and data
1.4.1 Database
In this chapter, as above, the source of information is microdata from the
Structure of Earnings Survey (SES) because it would be the dataset that comes
closest to our needs. SES is four year survey (1995, 2002, 2006 and 2010)
which offers independent cross-sectional datasets and includes matched
employer-employee microdata (observations for various workers employed in
each establishment). (Ramos R., Sanromà E., Simon H., 2014). We’ve obtained
the access to the microdata SES from Eurostat for different waves (2002, 2006
and 2010) and for all European countries7.
The EES offers microeconomic information about wages of an important
number of workers, about their characteristics, about workplace characteristics
and about establishment characteristics.
1.4.2 Countries and Collective bargaining regimes
From the SES database, we have chosen six different European countries,
according to available information for the breakdown “Annual bonuses” in
order to obtain one country, at least, for every bargaining regime. That way,
for Finland, Spain, Portugal, France, Romania and Poland the SES 2002 offers
separate information about different types of bonuses: regular bonuses,
productivity bonuses and profit sharing bonuses. This breakdown of “Annual
bonuses” is only available for the SES 2002.8 But we decided to keep the same
selection of six countries to be able to do comparative analysis between SES
2002, SES 2006 and SES 2010.
Moreover, these six countries account for different bargaining regimes in line
with the ICTWSS-Eurofound classification.
In order to take information about bargaining regimes, information from
variable “Collective pay agreement” from Structure of Earnings Survey (SES)
7 Eurostat-Research Proposal 53/2015-SES)
8 This breakdown of “Annual bonuses” in SES 2002 is based on the former version of the EU Regulation (1916/2000). From the SES 2006 onwards the EU Regulation (1738/2005) is implemented and this breakdown is no longer available.
9
data set is used. This variable indentifies the type of agreement covering at
least 50% of the employees in the local unit. The different options that are
offered by this variable are:
-National level or interconfederal agreement
-Industry agreement
-Agreement for individual industries in individual regions
-Enterprise or single employer agreement
-Agreement applying only to workers in the local unit
-Any other type of agreement
-No collective agreement exists
ICTWSS 4.0 is a time series dataset drawn up by J.Visser and hosted by the
Amsterdam Institute for Advanced Labour Studies (AIAS) which shows a
large collection of variables and indicators in Industrial Relations for EU and
OECD members (Eurofound, 2014).
We can combine this CA (collective agreement) SES classification with
ICTWSS 4.0 database (Visser J., 2013) and Eurofound classification
(Eurofound,2014) in Table 2:
10
Table 2. Bargaining regimes
ICTWSS code and description* Eurofound classification
SESclassification
5 = bargaining predominantly takes place at central or cross-
industry level
rather centralized
FINLAND-2002
FINLAND-2006 National CA level
4 = alternating between central and industry bargaining
rather centralized
SPAIN-2002
SPAIN-2006
SPAIN-2010
3 = bargaining predominantly takes place at the sector or
industry level
Intermediate
PORTUGAL-2002
PORTUGAL-2006
PORTUGAL-2010
FINLAND-2010
ROMANIA-2010
Industry CA level /Individual industries in individual regions CA
level
2 = alternating between sector and company bargaining
Intermediate
FRANCE-2002
FRANCE-2006
FRANCE-2010
ROMANIA-2002
ROMANIA-2006
1 = bargaining predominantly takes place at the local or
company level
rather decentralized
POLAND-2002
POLAND-2006
POLAND-2010
Enterprise or single employer CA/ Local
unit CA level
Source: Own elaboration from Visser (2013), Data Base on Institutional Characteristics of Trade Unions, Wage Setting, State Intervention and Social Pacts, 1960-2011 (ICTWSS) and SES (2006) Coding for categorical (or alphanumeric) variables. *We use the variable “level” which means the predominant level at which wage bargaining takes place. A bargaining level is “predominant” if it accounts for at least two-thirds of the total bargaining coverage rate in a given year and country. If it accounts for less, but for more than one-third of the coverage rate, there is a mixed or intermediate situation, between two levels. (Visser, 2013).
If we take the definition from Third European Company Survey, “Variable
Pay” refers to different components of pay which can vary over time in their
amount. In this way, a distinction is made between performance-related pay
which is linked to the performance of worker or group or workers and
financial participation which is linked to the company results (profit-sharing
schemes or employee share ownership schemes) (Eurofound, 2015).
11
1.4.3 Model wage determination
Our starting point is a Mincerian semilogarithmic wage equation (Mincer,
1974), where wages are determined by variables connected with human capital
(HC).
Where Wi is the natural logarithm of the gross hourly wage of worker i. HCi is a
vector of individual and capital human variables of worker i, including dummies
variables. So, HCi will be collecting labor supply factors.
But, in an extended way, we can include other variables (Palacio J.I, Simón H,
2004) (Lemieux, T. 2007) much more connected with labor demand factors. For
example: workplace factors (WP), establishment factors (ES), fixed effects for
every establishment (FE) and a proxy variable for Variable Pay schemes (VP) .
And taking into account all these aspects, we can obtain the following equation
(Simón H, 2009) (Marsden D, 1999):
Wij: natural logarithm of the gross hourly wage for each worker i HCi: vector of human capital variables for each worker i. Dummy variables. ESi: vector of variables describing establishment for every worker i . Dummy
variables WPi: vector of variables describing workplace for each worker i . Dummy
variables
θj: fixed effects for every establishment j
VPi : natural logarithm of the hourly bonuses for each worker i. This is a proxy
variable describing Variable Pay or Pay for Performance schemes
α:intercep
β,δ, γ, φ: vectors of parametres to be estimated
εij: random disturbance term
The wage equation (1) shows that we are in front of a multiple regression
analysis (several independent variables), with linear relationship among
parameters, where coefficients gives information about the change in
dependent variable for a 1 unit change in the predictor, holding other factors
12
fixed (ceteris paribus). As a regression methodology we’ve used OLS (ordinary
least squares) in order to analyze which are the main variables influencing
wage determination.
Our dependent variable is Gross Hourly Wage (natural logarithm). It is
calculated dividing the gross annual wage by annual agreed working hours
(both variables are from SES dataset).
Gross Hourly Wage =Gross Annual Wage9 / agreed annual workday
(lnhlyearning)
Independent variables:
*Individual factors (dummy variables with the exception of seniority): gender,
education, age, seniority.
*Workplace factors (dummy variables): occupation10, workday, contract,
supervisory
*Establishment factors (dummy variables): NACE11 classification, size,
control, market, collective agreement
And, as an additional independent variable we take into account a proxy
variable for Variable Pay or Pay for performance schemes and their
breakdowns. Hourly Annual Bonuses (natural logarithm):
For SES2002, SES2006 and SES2010
Hourly Annual Bonuses = Annual Bonuses12 / agreed annual workday
(lnhlybonuses)
Only for SES2002
Hourly Regular Bonuses = Regular Bonuses13 / agreed annual workday
(lnhlyregulbon)
9 Gross Annual Wage:total monetary remuneration received by workers during 2002,2006 and 2010 respectively Gross Anuual Wage includes = base pay + complements wage + withholding taxes + Special Variables Bonuses 10 Following ISCO-88 (COM). 11 Following NACE rev.1.1. 12 Includes any periodic, irregular, ad-hoc and exceptional bonuses and other payments that 13 Holiday bonuses, 13th and 14th month payment, allowances not taken and occasional commissions
13
Hourly Productivity Bonuses = Productivity Bonuses14 / agreed annual
workday
(lnhlyproductbon)
Hourly Profit sharing = Profit Sharing premiums15 / agreed annual workday
(lnhlyprofitsbon)
In order to analyze the influence of different factors and bonuses in wages, we
used a technique based on decomposition of the variance of individual wages.
This technique implies the estimation of different wage equation specifications
and the quantification of the variability in individual wages attributed to
different factors, through changes in determination coefficient. Marginal
contribution of each factor in the explanation of individual wage variability
measures associated effect of this factor. (Palacio J.I., Simón H., 2004). And
we applied this scheme to quantify the variability in individual wages attributed
to bonuses.16
We’ve called the different specifications as model A, model B and model C.
In model A, we analyze which are the most important factors determining
wages, controlling for human capital variables (gender, age, studies, seniority,
occupation) and for variable pay schemes variable.
In model B, we analyzed which are the most important factors determining
wages, controlling for human capital variables, workplace variables
(occupation, workplace, contract, responsibility), establishment variables
(Nace, size, market, regulation, ownership) and variable pay schemes variable.
In model C, we analyzed which are the most important factors determining
wages controlling for human capital variables, workplace variables and fixed
effects for establishments.
Effects establishments are not common in wage determination standard
models and were used as a novelty by Palacio J.I. and Simon H (Palacio and
Simon, 2004). These effects capture the impact on wages of the factors related
14 Bonuses linked to individual performance or piecework 15 Bonuses linked to the overall performance to the enterprise, under incentive schemes 16 An alternative approach to the influence of factors (especially demand factors) in the wage determination, could be the standard deviation of the establishment fixed effects, estimated from the full specification wage equation. This deviation is a measure of wage differentiation between establishments for workers with the same observable productive characteristics (Palacio J.I., Simón H., 2004).
14
to demand and they are used to control the heterogeneity17 between
establishments in wage determination. They could be analyzed through fixed
effects or through random effects. (Palacio and Simon, 2004). Hausman test
for 3 different SES waves and for 6 different countries indicates that these
effects are correlated with other explanatory variables. (Hausman, 1978). For
this reason, we’ve used fixed effects, because inappropriate use of random
effects supposes inconsistent estimation of the equation parameters (Hsiao,
1985). These effects must be considered representative of the sample but not
the entire population. (Greene, 1997).
As we explained before, in our dataset, we got information for 3 different SES
waves for each country: 2002, 2006 and 2010. So we’ve got 3 cross-section
independent datasets18 but not any panel data with the observations for the
same individuals through the time (Wooldridge, 2002). In this way is not
possible to separate the part of establishment effects due to unobserved
individual heterogeneity of obeying unobserved heterogeneity between the
establishments (Palacio JI and Simon H., 2004): we can only estimate global
effects. However, although control for unobservable individual fixed effects
tends to reduce the magnitude of wage differentials between establishments,
they persist significantly (Goux and Maurin, 1999) (Abowd et al, 1999)
(Abowd et al, 2001).1.5 Results
1.5.1 Wage determination results and explanation of wage variance
through R2 analysis
Through OLS regression, we analyzed wage determination for every of the 6
countries and for every SES wave, following 3 different models or
specifications:
ModelA: OLS regression. HC variables
ModelB: OLS regression. HC+WP+ES variables
ModelC: Fixed effect regression. HC+WP+fixed effects establishment
17 Differences across studied units 18 We could go further and analyze our data set as a pooled of independent cross sections,
introducing a dummy variable for every year (2002, 2006 and 2010).
15
The R-squared or R2 (coefficent of determination) gives us information about
level of regression fit to the data. But also, it gives us information about the
proportion of variance in the dependent variables which can be explained by
the independents variables (Wooldridge, 2002a).
In our case, we’ve used decomposition of wage variance through difference of
the R2 coefficient to see which part of this wage variance is explained by
bonuses. (Palacio J.I., Simón H. , 2004).
Difference R2-R2wab (without all bonuses) = contribution of bonuses to wage
variance
Difference R2-R2wrb (without regular bonuses) = contribution of regular
bonuses to wage variance
Difference R2wrb (without regular bonuses)-R2wab (without all bonuses) =
contribution of productivity bonuses and profit sharing premiums to wage
variance
16
1.5.1.1 Results for the SES 2002
Table 3. SES 2002 Annual Bonuses Model A, Model B and Model C
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
ANNUAL BONUSES VARIABLE PAY
SCHEMES lnhlybonuses MODEL A
0.271*** 0.396*** 0.677*** 0.271*** 0.276*** 0.521***
(0.00201) (0.00172) (0.00215) (0.00182) (0.00259) (0.00308)
R2 0.489 0.689 0.892 0.640 0.585 0.743
R2 wab19 0.25 0.47 0.48 0.39 0.31 0.40
Difference R2- R2 wab 0.239 0.219 0.412 0.25 0.275 0.343
lnhlybonuses MODEL B
0.242*** 0.336*** 0.637*** 0.229*** 0.224*** 0.452***
(0.00233) (0.00183) (0.00267) (0.00199) (0.00259) (0.00274)
R2 0.568 0.750 0.90620 0.715 0.661 0.839
R2 wab 0.41 0.62 0.6221 0.56 0.46 0.58
Difference R2- R2 wab
0.158
0.13
0.286
0.155
0.201
0.259
lnhlybonuses MODEL C
0.356*** 0.662*** 0.285*** 0.257*** 0.462***
(0.00364) (0.00711) (0.00356) (0.0107) (0.00827)
R2 0.643 0.84322 0.681 0.608 0.875
R2 wab 0.45 0.3223 0.48 0.40 0.56
Difference R2- R2 wab 0.193 0.523 0.201 0.208 0.315
Source: own elaboration from SES dataset
19 R2 calculated without all bonuses 20 Without occupation 21 Without occupation 22 Without occupation 23 Without occupation
17
Table 4. SES 2002 Regular Bonuses, Productivity Bonuses and Profit Sharing
Premium Model A
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
REGULAR BONUSES
PRODUCTIVITY BONUSES
PROFIT SHARING PREMIUMS MODEL A
lnhlyregulbon MODEL A
0.308*** 0.357*** 0.742*** 0.200*** 0.181***
(0.00415) (0.00413) (0.0182) (0.00533) (0.0180)
lnhlyproductbon MODEL A
0.0899*** 0.100*** 0.0468*** 0.152*** 0.201***
(0.00125) (0.00145) (0.00530) (0.00258) (0.0215)
lnhlyprofitsbon MODEL A
0.0807*** 0.0741*** 0.145***
(0.00953) (0.00291) (0.0150)
R2 0.645 0.741 0.953 0.720 0.871
R2 wrb 0.46 0.60 0.6824 0.679 0.79
R2 wab 0.25 0.47 0.4825 0.39 0.31
Difference R2- R2 wrb
0.185 0.141 0.273 0.041 0.081
Difference R2wrb- R2 wab
0.21 0.13 0.2 0.289 0.48
(0.0178) (0.0110) (0.0429)
Table 5. SES 2002 Regular Bonuses, Productivity Bonuses and Profit Sharing
Premium Model B
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
REGULAR BONUSES
PRODUCTIVITY BONUSES
PROFIT SHARING PREMIUMS MODEL B
lnhlyregulbon MODEL B
0.295*** 0.303*** 0.745*** 0.177*** 0.173***
(0.00496) (0.00463) (0.0174) (0.00551) (0.0190)
lnhlyproductbon MODEL B
0.0725*** 0.0964**
* 0.0450*** 0.126*** 0.190***
(0.00131) (0.00141) (0.00489) (0.00275) (0.0233)
lnhlyprofitsbon MODEL B
0.0808*** 0.0947*** 0.136***
(0.0107) (0.00301) (0.0143)
R2 0.714 0.797 0.95526 0.785 0.88627
24 Without NACE and without occupation 25 Without NACE and without occupation 26 Without NACE and without occupation
27 Without NACE and without occupation
18
R2 wrb28 0.5129 0.72 0.7330 0.7031 0.8432
R2 wab33 0.41 0.62 0.5834 0.55 0.46
Difference R2- R2 wrb
0.204 0.077 0.225 0.085 0.046
Difference R2wrb- R2 wab
0.1 0.1 0.15 0.15 0.38
Table 6. SES 2002 Regular Bonuses, Productivity Bonuses and Profit Sharing
Premium Model C
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
REGULAR BONUSES
PRODUCTIVITY BONUSES
PROFIT SHARING PREMIUMS MODEL C
lnhlyregulbon MODEL C
0.339*** 0.711*** 0.325*** 0.223***
(0.0102) (0.0476) (0.0193) (0.0612)
lnhlyproductbon MODEL C
0.123*** 0.0363*** 0.118*** 0.219***
(0.00255) (0.00629) (0.00483) (0.0412)
lnhlyprofitsbon MODEL C
0.106*** 0.219*** 0.180***
(0.0178) (0.0110) (0.0429)
R2 0.70435 0.93336 0.800 0.82337
R2 wrb38 0.61 0.6339 0.7040 0.8041
R2 wab42 0.45 0.3243 0.479 0.40
Difference R2- R2 wrb
0.094 0.303 0.1 0.023
Difference R2wrb- R2 wab
0.16 0.31 0.221 0.4
28 Without regular bonuses 29 Without occupation 30 Without NACE and without occupation 31 Without occupation
32 Without occupation and NACE
33 Without all bonuses 34 Without NACE and without occupation 35 Without seniority 36 Without occupation 37 Without age, occupation, workday, contract and supervisory 38 Without regular bonuses 39 Without occupation 40 Without occupation 41 Without occupation 42 Without all bonuses 43 Without occupation
19
Tables 3, 4, 5 and 6 summarize the main results about the incidence of
bonuses in wage determination, from OLS regression using SES 2002 dataset.
Through three different models, in six European countries, we’ve analyzed the
incidence of annual bonuses and the incidence of his breakdown in regular
bonuses, productivity bonuses and profit sharing.
In the case of annual bonuses for model A (taking to account only individual
factors and variable pay), we can explain that Portugal and Poland have higher
incidence in wage determination in comparison with the rest of countries.
These two countries have intermediate and decentralized collective bargaining,
respectively. They show that for every increase of 1% in annual bonuses wage
increases is 0.67% in the case of Portugal and is 0.52% in the case of Poland.
Also, we can say that in these 2 countries the proportion of wage variance
explained by individual factors and variable pay are higher than the rest of
countries: 89.2% and 74.3%. And if we analyze the incidence in wage variance
of bonuses, comparing R2 with bonuses with R2 without bonuses, we can say
that in Portugal and Poland bonuses explain the main part of wage variance
respect to the rest of countries: 41.2% in the case of Portugal and 34.3% in the
case of Poland. For the rest of countries, bonuses explain about 20% of wage
variance.
In the case of model B, (taking to account all factors: individual factors,
workplace factors, establishment factors and variable pay), again in Portugal
and Poland annual bonuses have higher incidence in wage determination: for
every increase of 1% in annual bonuses, wage increases is 0.63% in the case of
Portugal and is 0.45% in the case of Poland. And in these 2 countries, bonuses
explain respectively 28.6% of wage variance and 25.9% of wage variance.
Same explanation can be used for model C (including individual factors,
workplace factors and fixed effects for establishments), where Portugal
explains 52.3% of wage variance and Poland explains 31.5% of wage variance.
In these 2 countries, bonuses are so important that in Portugal and with model
C, they are explaining more than 50% of wage variance and in Poland, with
model A, they are explaining 34.3%.
Now, if we have a look to the breakdown of annual bonuses, in model A and
in the case of regular bonuses, we can say that Portugal and Spain are
countries with higher incidence in wage determination: for every increase of
1% in regular bonuses, wage increases are 0.74% and 0.35% respectively. In
the case of productivity bonuses, Romania and France have the first and
20
second position, respectively: for every increase of 1% in productivity
bonuses, wage increases are 0.20% and 0.15%. And finally, Romania is the
country with higher incidence of profit sharing in wage determination: for
every increase of 1% in profit sharing, wage increases are 0.14%.
Regular bonuses explain 27.3% of wage variance for Portugal and 18.5% of
wage variance for Finland, as the second country with a higher explanation.
Spain is the third country with an explanation of 14.1% of wage variance. This
percentage is lower for France (4.1%) and for Romania (8.1%). So, countries
with more centralized collective bargaining have higher percentage of
explanation of wage variance by regular bonuses and countries with more
decentralized collective bargaining have lower percentage. And, on the
opposite side, countries with more centralized collective bargaining have lower
percentage of explanation of wage variance by productivity bonuses and profit
sharing and countries with more decentralized collective bargaining have
higher percentage: 48% for Romania and 28.9% for France.
In model B, Portugal and Spain are countries with higher incidence of regular
bonuses in wage determination (0.74% and 0.30%, respectively). Poland and
Romania are countries with bigger incidence of productivity bonuses in wage
determination (0.19% and 0.12% respectively) and with bigger incidence of
profit sharing in wage determination (0.13% and 0.09%, respectively).
Regular bonuses explain 22% and 20% of wage variance for Portugal and
Finland. And productivity bonuses and profit sharing explain 38% of wage
variance for Poland and the 15% for Romania and Portugal. So, as with model
A, countries with more centralized collective bargaining have higher
percentage of wage variance explanation by regular bonuses and countries
with more decentralized collective bargaining have higher percentage of
explanations of wage variance by productivity bonuses and profit sharing.-
In model C, Portugal and Spain again are countries with higher incidence of
regular bonuses in wage determination (0.71% and 0.33% respectively). Poland
and Spain are countries with bigger incidence of productivity bonuses in wage
determination (0.21% and 0.12% respectively). And Romania and Poland are
countries with bigger incidence of profit sharing in wage determination (0.21%
and 0.18%) respectively. Regular bonuses explain 30% and 10% for Portugal
and France respectively. Productivity bonuses and profit sharing explain 40%
of wage variance for Poland and 31% of wage variance for Portugal. Unlike
model A and in model B, countries with more decentralized collective
21
bargaining have higher percentage of wage variance explanation by regular
bonuses and by productivity bonuses and profit sharing.
1.5.1.2 Results for the SES 2006
Table 7. SES 2006 Model A Model B and Model C
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
ANNUAL BONUSES VARIABLE PAY
SCHEMES lnhlybonuses MODEL A
0.171*** 0.378*** 0.127*** 0.244*** 0.876***
(0.00121) (0.00178) (0.00129) (0.00131) (0.00241)
R2 0.452 0.678 0.508 0.580 0.921
R2 wab44 0.31 0.43 0.359 0.37 0.33
Difference R2- R2 wab 0.142 0.248 0.149 0.21 0.591
lnhlybonuses MODEL B
0.140*** 0.328*** 0.118*** 0.212*** 0.834***
(0.00127) (0.00185) (0.00141) (0.00127) (0.00327)
R2 0.575 0.737 0.591 0.65945 0.945
R2 wab46 0.49 0.57 0.51 0.51 0.50
Difference R2- R2 wab 0.085 0.167 0.081 0.149 0.445
lnhlybonuses MODEL C
0.344*** 0.151*** 0.264*** 0.920***
(0.00368) (0.00326) (0.00500) (0.00513)
R2 0.642 0.549 0.67447 0.971
R2 wab48 0.41 0.44 0.48849 0.41
Difference R2- R2 wab 0.232 0.109 0.186 0.561
Source: own elaboration from SES dataset
In table 7, we analyzed same results schemes like tables 3, 4, 5 and 6, but for
the SES 2006 dataset. In this case, we don’t have any breakdown about annual
bonuses. Poland and Spain show the highest level of incidence in wage
determination in model A: for every increase of 1% in bonuses, wages increase
0.87% for Poland and 0.37% for Spain. Bonuses would explain 59% of wage
variance in the case of Poland and the 24.8% for Spain.
44 R2 calculated without all bonuses 45 Without contract 46 R2 calculated without all bonuses 47 Without contract 48 R2 calculated without all bonuses 49 Without contract
22
In model B and C, Poland and Spain show highest incidence in wage
determination: 0.83% and 0.92% for Poland and 0.32% and 0.34% for Spain.
In model B, bonuses would explain 44.5% of wage variance for Poland and
16.7% of wage variance for Spain. In model C, bonuses would explain 56.1%
of wage variance for Poland and 23.2% of wage variance for Spain.
1.5.1.3 Results for the SES 2010
Table 8. SES 2010 Model A Model B and Model C
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
ANNUAL BONUSES VARIABLE PAY
SCHEMES lnhlybonuses MODEL A
0.0722*** 0.561*** 0.164*** 0.283*** 0.172***
(0.00124) (0.00240) (0.00117) (0.00154) (0.000680)
R2 0.169 0.741 0.494 0.532 0.570
R2 wab50 0.15 0.25 0.34 0.35 0.43
Difference R2- R2 wab 0.019 0.491 0.154 0.182 0.14
lnhlybonuses MODEL B
0.0530*** 0.470*** 0.141*** 0.251*** 0.144***
(0.00122) (0.00260) (0.00118) (0.00138) (0.000625)
R2 0.259 0.789 0.637 0.63551 0.724
R2 wab52 0.253 0.37 0.54 0.517 0.63
Difference R2- R2 wab 0.006 0.419 0.097 0.118 0.094
lnhlybonuses MODEL C
0.578*** 0.191*** 0.292*** 0.207***
(0.00498) (0.00301) (0.00534) (0.00658)
R2 0.771 0.624 0.63653 0.752
R2 wab54 0.295 0.50 0.46255 0.37256 Difference R2- R2 wab 0.476 0.124 0.174 0.38
Source: own elaboration
In table 8, we analyzed same results schemes like table 3 and table 7 but for
the SES 2010 dataset. As above, we don’t have any breakdown about annual
bonuses. Spain and Romania show the highest level of incidence in wage
50 R2 calculated without all bonuses 51 No occupation 52 R2 calculated without all bonuses 53 No occupation 54 R2 calculated without all bonuses 55 No occupation 56 No occupation
23
determination in model A: for every increase of 1% in bonuses, wages increase
0.56% for Spain and 0.28% for Romania. Bonuses would explain 49% of wage
variance in the case of Spain and the 18.2% for Romania. Same results could
be got from model B and model C. In model B, bonuses would explain 41%
of wage variance for Spain and 11.8% for Romania. In model C, bonuses
would explain 47.6% of wage variance for Spain and 17.4% for Romania.
Table 9. Summary SES 2002, 2006 and 2010 Model A, B and C
COEFFICIENTS FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
ANNUAL BONUSES VARIABLE PAY
SCHEMES
SES 2002
Difference R2- R2 wab MODEL A
0.239 0.219 0.412 0.25 0.275 0.343
Difference R2- R2 wab MODEL B
0.158
0.13
0.286
0.155
0.201
0.259
Difference R2- R2 wab MODEL C
0.193 0.523 0.201 0.208 0.315
SES 2006
Difference R2- R2 wab MODEL A
0.142 0.248 0.149 0.21 0.591
Difference R2- R2 wab MODEL B
0.085 0.167 0.081 0.149 0.445
Difference R2- R2 wab MODEL C
0.232 0.109 0.186 0.561
SES 2010
Difference R2- R2 wab MODEL A
0.019 0.491 0.154 0.182 0.14
Difference R2- R2 wab MODEL B
0.006 0.419 0.097 0.118 0.094
Difference R2- R2 wab MODEL C
0.476 0.124 0.174 0.38
Source: own elaboration from SES data set
Looking the evolution of explanation of wage variance by bonuses from 2002
SES dataset to 2010 SES data set (table 9), we can say that in most of six
countries its percentage has been decreasing. The exception is Spain where
this percentage has been increasing close to 50%. In other countries, like
Romania this percentage has been reduced.
The explanation of this evolution could be that some countries are witnessing
a decentralization process in their collective bargaining. This could means that
the part of regular bonuses from total annual bonuses is decreasing in front of
the part of productivity bonuses and profit sharing premiums.
24
As we found in previous chapter, due to Finland change it collective
bargaining level becoming less decentralized, Spain is the country with higher
level of collective bargaining. And Romania, which had a decentralized level of
collective bargaining, from 2010 it become much more centralized. So, in spite
of this evolution, the final results in SES 2010 show that in countries with
higher level of centralization in collective bargaining are those countries with
higher percentage in wage variance explanation by bonuses. This could mean
that, in those countries, weight of regular bonuses in total annual bonuses is
higher than in the rest of countries. But to go further in this conclusion we
would need breakdown of bonuses in SES 2006 and in SES 2010.
1.5.2 More inequality analysis: Gini Index and Fields decomposition
1.5.2.1 Gini Index and variance of logarithms
If we want to evaluate one dimension of inequality like income inequality
applied to wages, we could use the evolution of Gini Index and variance
logarithms, as dispersion measures. With our dataset, we compared results for
the SES wave 2002, SES wave 2006 and the SES wave 2010.
In tables 10, 11 and 12, we analyzed Gini Index and Variance log of gross
annual salary, bonuses and of a proxy of base pay57, which is calculated
subtracting bonuses for gross annual salary. Because could be interesting to
understand the dispersion level taking to account bonuses and without taking
to account in whole annual salary.
57 SES not offers base pay information for every country. We have to remind that all bonuses
includes regular bonuses, productivity bonuses and profit sharing premiums
25
Table 10. SES 2002 Gini Index and Variance log
2002 year
FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
Proxy Base
Pay
Gini index 0.22 0.3 0.35 0.35 0.4 0.34
Variance
log 0.14 0.26 0.32 0.38 0.42 0.32
Gross
Annual Salary
Gini index 0.23 0.31 0.36 0.37 0.42 0.34
Variance
log 0.14 0.27 0.35 0.38 0.41 0.34
All Bonuses
Gini index 0.43 0.43 0.46 0.61 0.62 0.33
Variance
log 0.62 0.55 0.56 1.39 1.35 0.40
Regular bonuses
Gini index 0.32 0.37 0.38 0.44 0.58 0.3
Variance
log 0.41 0.45 0.40 0.79 1.17 0.32
Productivity
bonuses
Gini index 0.58 0.63 0.61 0.75 0.71
Variance
log 1.39 1.82 1.61 2.25 1.56
Profit
Sharing premiums
Gini index
0,49
0.76 0.43
Variance
log 1,04
2.02 0.66
Source: own elaboration from SES dataset
From table 10, we can say that Finland is the country with more equal
distribution in its gross annual salary and its proxy of base pay, according to a
Gini Index closer to 0 (0.22 in the case of proxy base pay and 0.23 in the case
of gross annual salary). On the contrary, Romania would become the country
with less equal distribution, according to a Gini Index of 0.4 for proxy base
pay and 0.42 for gross annual salary. The same scheme could be found in the
case of variance of logarithms.
26
Looking into bonuses, we can observe, in general, higher Gini Index and
higher variance of logarithms in all countries in comparison with base pay and
gross annual salary. In the case of all bonuses and regular bonuses, Romania is
again the country with higher Gini Index, with levels of 0.62 and 0.58,
respectively. And Poland is the country with lower Gini Index, with levels of
0.3 and 0.33 respectively. For Productivity bonuses, Finland has lowest Gini
Index (0.58) and France has highest Gini Index (0.75). And finally, in the case
of profit sharing premium, again Romania has highest Gini Index (0.76) and
Poland lowest Gini Index (0.43).
That is our departure situation and we are going to compare these results with
situation in 2006 and 2010.
Table 11. SES 2006 Gini Index and Variance log
2006 year FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
Proxy Base Pay
Gini index
0.24 0.29 0.38 0.28 0.41 0.41
Variance log
0.16 0.2401 0.4225 0.2209 0.4624 0.4225
Gross Annual Salary
Gini index
0.24 0.3 0.38 0.3 0.42 0.41
Variance log
0.16 0.2601 0.4356 0.2401 0.49 0.4225
All Bonuses
Gini index
0.46 0.43 0.43 0.69 0.61 0.42
Variance log
0.7744 0.5929 0.5329 2.3409 1.5625 0.5041
Source: own elaboration from SES dataset
Table 11, shows that in 2006, in the case of proxy base pay and gross annual
salary, Finland continues being the country with more equal distribution,
because it has lower Gini Index in comparison with the rest of countries. In
the same way, Romania continues being the country with less equal
distribution. But the main differences with 2002 results are that both Finland
and Romania don’t highlight in the same way. For example, close to Finland,
other countries like Spain of France also show not very high Gini Index. And
close to Romania, Poland shows a high Gini Index. So, in the case of proxy of
base pay and in the case of gross annual salary, we can find 2 groups of
27
countries. One group with higher Gini Index level and inequality (Romania,
Poland and Portugal) and other group with lower Gini Index level (Finland,
France and Spain).
If we have a look to bonuses (without breakdown), as in 2002, Poland has
lower Gini Index (0.42) and France has higher Gini Index (0.69). Poland is not
alone in its position, because it is followed very close by Spain and Portugal.
Table 12. SES 2010 Gini Index and Variance log
2010 year
FINLAND SPAIN PORTUGAL FRANCE ROMANIA POLAND
Proxy Base Pay
Gini index 0.24 0.4 0.38 0.27 0.4 0.33
Variance
log 0.1681 0.5776 0.3969 0.2025 0.4356 0.3136
Gross Annual Salary
Gini index 0.4 0.4 0.39 0.29 0.41 0.33
Variance
log 0.7056 0.5776 0.4096 0.2209 0.4489 0.3249
All Bonuses
Gini index 0.6 0.52 0.48 0.59 0.6 0.49
Variance
log 2.3409 0.9801 0.6241 1.3689 1.3225 1.2769
Source: own elaboration from SES dataset
In Table 12 results from the SES 2010 could be found. Here, we have to
mention some relevant changes respect to previous tables. In the case for
proxy of base pay, Finland is the most equal country, yet. But, Spain, has
worsened its situation, because from a Gini Index of 0.3 in 2002 and of 0.29 in
2006, it show a higher Gini Index of 0.4 in 2010, reaching the level of
Romania. But if we compare the situation in proxy of base pay with the
situation in gross annual salary, we can see that, apart from Spain, surprisingly,
Finland has deteriorated its position, reaching also Romania level. France
remained as the country with more equal distribution of its gross annual salary.
This explanation is connected with the fact that, in the case of bonuses,
Finland and Spain are showing a higher Gini Index of 0.4 close to Romania
level. So, one of the reasons that could give an explanation to the deterioration
of the Finland position in gross annual salary is its deterioration in bonuses
equality level. And this fact, like the changes in Spain, could be caused by the
28
effects of the international economic crisis. But, in the case of Finland, this
situation could be related to the change in collective bargaining regime, which
has been analyzed in previous chapter: as much more decentralized bargaining
more inequality in gross annual salary. Nevertheless, if this argument was true,
we have to expect that Romania will improve its position, because their
collective bargaining regime changed, becoming more centralized and that is
not the case, because its position it is rather similar through in 2002, 2006 and
2010 dataset. We can say that changes in collective bargaining regime were
higher in Finland than in Romania: Finland changes from rather centralized
(level 5) to intermediate (level 3) bargaining regime and Romania changes from
intermediate (level 2) to intermediate (level 3). But, it is difficult to find a clear
pattern.
1.5.2.2. Fields decomposition
Going further, if we want to know which are the weight of different factors in
wage inequality, we can use Fields decomposition58 Following this procedure,
the dispersion of dependent variable measured by variance, for example, is
broken down into a number of components such the whole is equal to the
sum of its parts. (Fields, 2003b).
We calculated Fields decomposition for all factors59 but we presented
summarized results pooled in different groups: factors connected with human
capital (HC), connected with workplace (WP), connected with establishment
(ES) and bonuses.
In every table, we can observe Fields decomposition with information from
SES dataset, using proxy of base pay, total gross annual salary and total gross
annual salary plus bonuses. We are aware that in last case, as bonuses are
included in gross annual salary, in this decomposition reversal causation60
could be found. Because we are analyzing inequality of gross annual salary
with factorial Fields decomposition and bonuses is one factor which is inside
gross annual salary at the same time. So, when we are analyzing inequality of
58 We’ve used ineqrbd stata instruction designed by Fioro and Jenkins (2007)
59 We could go further and also have done the same calculations taking to account only
human capital factors, human capital plus workplace factors and plus fixed effects of
establishment. Like in (Simon H., 2009)
60 In this sense, inequality in gross annual salary are determining inequality in bonuses or
inequality in bonuses are determining inequality in gross annual salary.
29
gross annual salary, we are analyzing inequality in bonuses, too. But, we are
interested in look into which part of inequality of gross annual salary is
explained by different factors and by bonuses. For this reason, we’ve decided
calculate the same decomposition for gross annual salary without bonuses
(proxy Base pay), gross annual salary (with bonuses) and gross annual salary
with breakdown for bonuses and we compare all the results.
30
Table 13. SES 2002 Fields decomposition
FIELDS DECOMPOSITION
YEAR 2002
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
FINLAND FINLAND FINLAND SPAIN SPAIN SPAIN PORTUGAL PORTUGAL PORTUGAL
HC factors 14.06 15.04 8.82 21.59 22.56 13.52 18.53 18.77 5.01
WP factors 21.01 21.47 14.20 26.60 27.27 16.98 25.45 25.71 7.16
ES factors 4.12 4.88 1.69 11.74 12.71 7.82 24.78 25.09 6.57
Bonuses
32.12
36.63
72.35
Total 39.18 41.38 56.83 59.93 62.55 74.96 68.76 69.57 91.09
Total-bonuses
24.71
38.32
18.73
Residual 60.82 58.62 43.17 40.07 37.45 25.04 31.24 30.43 8.91
FRANCE FRANCE FRANCE ROMANIA ROMANIA ROMANIA POLAND POLAND POLAND
HC factors 16.62 17.70 13.46 21.28 17.14 14.84 18.61 18.78 11.06
WP factors 31.39 34.42 25.00 19.08 16.10 15.25 31.08 31.39 29.64
ES factors 2.28 3.79 1.08 12.77 13.36 7.15 8.01 8.08 0.73
Bonuses
31.92
28.87
42.45
Total 50.28 55.91 71.46 53.13 46.60 66.11 57.71 58.25 83.87
Total-bonuses
39.54
37.24
41.42
residual 49.72 44.09 28.54 47.05 53.51 33.90 42.29 41.75 16.13
Source: own elaboration
31
In table 13, results of Fields decomposition with the SES 2002 dataset61 are
showed. In general, we can say that important factors which are playing an
important role in inequality are gender (especially in the case of Finland and
Spain), some levels of education (Romania, Poland, Portugal and France),
seniority (Spain), some kinds of occupation (France, Poland, Portugal, Spain),
some kinds of sectors (Portugal), companies with more than 250 workers
(Spain, Portugal and Romania) and collective agreement at enterprise level
(Spain and Portugal).
If we compare the situation for proxy base pay, we can say that in most of all
of six countries workplace factors are determining the higher percentage of
inequality. The exception would be Romania, where education has a bigger
importance in inequality in comparison with the rest of countries.
All the considered factors are explaining the 68.76% of inequality in proxy of
base pay in Portugal and 59.93% of inequality in proxy of base pay in Spain. If
we would take into account fixed effects of establishments we could observe
probably that establishment factors would have higher incidence in some
countries. Nevertheless, establishment factors have biggest influence for
Spain, Romania and especially for Portugal, explaining between 24% and 25%
of inequality. The importance of workplace elements is bigger in France and
Poland due to the role of some kind of occupations.
If we observe the Fields decomposition considering gross annual salary plus
bonuses, we can say that in the case of Portugal, Poland and Finland the
contribution of bonuses are higher than the sum of other factors. In Portugal,
bonuses would explain 72.35% of total inequality and the 18.73% the
remaining factors. In the case of Poland, bonuses would explain 42.45% of the
total inequality (83.87%) and the remaining factors the 41.42%, reaching 2
percentages almost to the same level. In Finland, bonuses would explain
32.12% of 56.83% total inequality and the rest of factors the 24.71%
remaining. So, taking into account that a part of the total inequality of gross
annual salary includes inequality of bonuses, we can say that in the case of
Portugal, Poland (with only a difference of 1 percentual point) and Finland
bonuses have high incidence in gross annual salary inequality in comparison
61 In appendices all details about Fields decomposition could be found
32
with the rest of factors62 . These results are consistent with the previous
section in which we have said that in Portugal and Poland bonuses explain the
main part of wage variance respect to the rest of countries and that in Finland
regular bonuses explain the main part of wage variance respect to the rest of
countries.
In the rest of countries, the contribution of bonuses to total gross annual
salary inequality is lower than the rest of elements, especially in the case of
France and Romania (in the case of Spain the quantities are very similar). So,
we can conclude that in these three countries, bonuses have highest (Portugal
and Poland) and high (Finland) level of incidence in inequality level and all
bonuses (Portugal and Poland) and regular bonuses (Finland) explain the main
part of wage variance.
62 If we have to talk about countries with highest incidence of bonuses in gross annual salary
inequality we have to refer to Portugal, Poland and Spain. But in the case of last country, the
incidence of bonuses is not higher than the rest of factors.
33
Table 14. SES 2006 Fields decomposition
FIELDS DECOMPOSITION
YEAR 2006
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
FINLAN
D FINLAND FINLAND SPAIN SPAIN SPAIN
PORTUGAL
PORTUGAL
PORTUGAL
HC factors 16.85 18.46 15.18 21.76 22.99 13.86 28.73 28.86 6.66
WP factors 24.32 23.86 19.68 23.33 23.86 15.02 32.86 33.03 8.04
ES factors 4.76 7.38 5.92 9.96 10.91 6.94 12.26 11.95 4.77
Bonuses
16.77
37.84
74.64
Total 45.93 49.70 57.54 55.05 57.76 73.66 73.84 73.84 94.10
Total-bonuses
40.77
35.82
19.46
residual 54.07 50.30 42.46 44.95 42.24 26.34 26.16 26.16 5.90
FRANCE FRANCE FRANCE ROMANIA ROMANIA ROMANIA POLAND POLAND POLAND
HC factors 22.20 21.30 18.83 16.39 16.19 14.29 14.23 14.33 2.25
WP factors 26.67 24.80 19.90 24.84 24.51 22.84 31.62 32.06 9.60
ES factors 3.16 5.17 1.47 9.08 10.41 2.35 3.63 4.00 -1.06
Bonuses
18.93
26.47
83.76
Total 52.03 51.27 59.13 50.31 51.11 65.95 49.48 50.39 94.55
Total-bonuses
40.20
39.48
10.79
residual 47.97 48.73 40.87 49.71 48.91 34.08 50.52 49.61 5.45
Source: own elaboration
34
Table 14 shows a summary of the main results of Fields decomposition from
SES 2006 dataset. The most important elements, in this case, which have a
higher incidence on inequality are gender (Finland and Spain), some high
levels of education, seniority (Spain and Portugal), some kinds of occupation
and some kinds of sectors. Like in results from SES 2002 dataset, workplace
factors are which have biggest influence in inequality of Proxy base pay and in
inequality of annual gross salary.
Considering all factors together, we can remark that they are explaining 55%
of Spain Proxy base pay and 73.84% of Portugal Proxy base pay. And we can
say that all factors are explaining 73.66% of gross annual salary plus bonuses
inequality in Spain, 94.10% in Portugal and 94.55% in Poland.
Analyzing Fields decomposition, if we consider gross annual salary plus
bonuses, we can say that in the case of Portugal, Poland and Spain the
contribution of bonuses are higher than the sum of other factors. In Portugal,
bonuses would explain 74.64% of total inequality and the 19.46% the
remaining factors. In the case of Poland, bonuses would explain 83.76% of the
total inequality (94.55%) and the remaining factors the 10.79%. In Spain case,
bonuses would explain 37.84% of 73.66% total inequality and the rest of the
factors would explain the 35.82% remaining. In this way, we can say that in
the case of Portugal, Poland and Spain bonuses have biggest incidence in gross
annual salary inequality in comparison with the rest of factors. These results
are coherent with the previous section in which we have said that, using SES
2006 dataset, in Poland and Spain bonuses explain the main part of wage
variance respect to the rest of countries63.
63 We don’t have results for Portugal regression
35
Table 15. SES 2010 Fields decomposition
FIELDS DECOMPOSITION
YEAR 2010
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
Proxy BASE PAY
GROSS ANNUAL SALARY
GROSS ANNUAL SALARY + BONUSES
FINLAND FINLAND FINLAND SPAIN SPAIN SPAIN PORTUGAL PORTUGAL PORTUGAL
HC factors 15.98 8.35 7.63 9.80 11.61 6.93 29.71 30.11 14.18
WP factors 26.57 12.81 12.27 21.20 23.32 18.53 27.67 28.02 14.00
ES factors 5.50 4.14 3.93 2.13 2.95 2.36 10.32 10.50 4.74
Bonuses
2,08
51,03
51,50
Total 48.05 25.30 25.91 33.13 37.89 78.86 67.70 68.63 84.42
Total-bonuses
23.83
27.82
32.92
Residual 51.95 74.70 74.09 66.87 62.11 21.14 32.30 31.37 15.58
FRANCE FRANCE FRANCE ROMANIA ROMANIA ROMANIA POLAND POLAND POLAND
HC factors 18.17 18.31 14.42 16.48 16.42 11.73 17.04 17.16 14.94
WP factors 29.96 28.61 24.64 27.33 26.93 21.38 42.25 41.42 41.32
ES factors 6.21 7.60 6.39 7.90 8.44 6.40 2.81 4.21 -1.05
Bonuses
18.26
26.55
17.03
Total 54.34 54.53 63.71 51.71 51.79 66.05 62.10 62.79 72.24
Total-bonuses
45.45
39.51
55.21
Residual 45.66 45.47 36.29 48.29 48.21 33.95 37.69 36.99 27.57
Source: own elaboration
36
Finally, in table 15 we included the main results from Fields decomposition using
SES 2010 dataset. Like in previous table, in most of countries, workplace factors are
those which have highest contribution to inequality. Portugal is the exception with
bigger percentages relevance of human capital factors respect to workplace factors
and establishment factors. The main reason for this pattern is the higher weight of
some education level (bachelor and master) respect to the rest of countries.
Other important aspects to highlight are that gender is not as important element in
inequality as in previous years (for Finland and Spain). Moreover, education is not as
important as in 2002 and 2006 dataset; the exception is Romania, Poland and
Portugal (as we’ve just explained). Seniority is important, not only for Spain, but for
Portugal and Poland. Some kinds of occupations have big importance in inequality,
partial workday is outstanding for Spain and supervisory for France. Size of
companies with more than 250 workers is an important element for Romania and
Portugal. And finally, collective agreement at enterprise level is important for
Portugal.
Again, low weight of establishment factors could be connected with the fact that
we are not taking into account fixed effects.
Taking to account gross annual salary plus bonuses, we can say that in the case of
Portugal, Spain and Romania the contribution of bonuses are higher than the sum
of other factors. In Portugal, bonuses would explain 51.5% of total inequality and
the 32.92% the remaining factors. In the case of Spain, bonuses would explain 51%
of the total inequality (78.8%) and the remaining factors the 27.8%. In Romania,
bonuses would explain 26.5% of 66% total inequality and the rest of factors the
39.5% remaining. In this way, we can say that in the case of Spain and Portugal
bonuses have biggest incidence in gross annual salary inequality in comparison with
the rest of factors. In the case of Romania bonuses have one the highest incidence
in gross annual salary inequality but this percentage in not superior to the incidence
of the rest of factors. Again, these results are connected with the results in previous
section in which we have said that, using SES 2010 dataset, in Spain and Romania,
bonuses are explaining the main part of wage variance respect to the rest of
countries64.
Finally, we’d like to remark the low incidence of bonuses in inequality of Finland of
only 2% in comparison with the rest of countries and in comparison with previous
years. One possible explanation of this fact could be that as Finland changes its
collective bargaining regimes, from more centralized to more decentralized, regular
64 We don’t have results for Portugal regression
37
bonuses would be losing important in front of the other types of bonuses. We have
to remind that in Finland regular bonuses were important factors explaining the
main part of wage variance.
1.6. Conclusions
The main objective of this paper is to show if the introduction and use of Variable
pay systems implies much more wage inequality. Using data from three waves of
SES (2002, 2006 and 2010), comparing six selected countries, and considering
bonuses as a proxy of Variable pay systems, we can say that there is some
relationship between bonuses and wage inequality.
Our results show that in five countries of our analyzed group, Portugal, Poland,
Spain, Romania and Finland (only in the case of regular bonuses), bonuses are
explaining the most important part of wage variance. And in these same five
countries bonuses would have highest incidence in gross annual salary inequality,
following Fields decomposition. With 2002 dataset relevant countries of these five
are Portugal, Poland and Finland. With 2006 dataset are Portugal, Poland and Spain.
And with 2010 dataset are Portugal, Spain and Romania. So, France would be the
only country which wouldn’t be affected by the wage inequality with the
introduction of bonuses.
If we analyze the evolution of bonuses contribution to wage variance, we can say
that this hasn’t been clearly increasing from 2002 to 2010, except for the case of
Spain. Despite this evolution, the final result is that countries with much more
centralized collective bargaining have higher contribution of bonuses in wage
variance, those countries with a less centralized collective bargaining and with less
weight of regular bonuses in total bonuses. This would be a good explanation for
the evolution of Finland, which has a much more decentralized collective
bargaining. But this would be true only in the case of regular bonuses. It would be
necessary the breakdown of bonuses for the SES 2006 and SES 2010, to get much
consistent results with literature.
Obviously, apart from bonuses, other factors could be influencing in wage
inequality of this countries. For example, structural factors or their starting point. If
we compare situation with the SES 2002 dataset with the situation with SES 2010
dataset, we can say that France as Poland (only slightly) have improved their
situation in terms of Gini Index, meanwhile Spain, Portugal have worsened their
situation. Romania has remained at highest level of Gini Index, between 2002 and
2010. Finland is a special case because it has worsened their situation only in the
case of gross annual salary Gini Index, but not for the case of proxy base pay Gini
Index.
38
So, we can conclude that Variable pay systems can imply a bigger deterioration of
wage inequality especially in countries with general wage inequality problems. This
could be the case for countries with Gini Index up 0.35 like Spain, Portugal and
Romania.
For some authors (Lemieux et al, 2007) if variable pay or pay for performance can
explain an important part of wages variance, it could happen that complementarities
in production may be less important than individual’s contribution to output.
But some other literature link wage inequality with skill-biased technical change
(Acemoglu, 2002). And in this case we have to analyze deeper complementarities in
production. However, variable forms of compensation can be understood as a form
of “technology” to adapt to new circumstances (Lemieux et al, 2007).
39
BIBLIOGRAPHY
Abowd, J.M. y F. Kramarz (1999): “The Analysis of Labor Markets Using Matched
Employer Employee-Data”, en O. Ashenfelter y D. Card (eds.) Handbook of
Labor Economics. ed. North-Holland.
Abowd, J.M., F. Kramarz, D.N. Margolis y K. Troske (2001): “The Relative
Importance of Employer and Employee Effects on Compensation: A
Comparison of France and the United States”, Journal of the Japanese and
International Economics, n.º 15, págs. 419-436.
Acemoglu, D. (2002, Mar). Technical change, inequality, and the labor market.
Journal of Economic Literature 40(1), 7:72.
Blinder, A. S. (1973). Wage Discrimination: Reduced Form and Structural
Estimates. Journal of Human Resources, 8, 436-455.
Bryson, Freeman, Lucifora et al (2010). “Paying for Performance”. XII European
Conference of the Fondazione RODOLFO DEBENEDETTI.
Cowell Frank A. and Fiorio Carlo V. (2010). “Inequality decompositions”. Gini
discussion paper 4. Growing inequalities’impacts.
ETUI (2012). Benchmarking working Europe 2012. “Collective bargaining and
diversity in wage developments”. Brussels.
Eurofound (2015). Third European Company Survey-2013 – Overview report:
Workplace practices – Patterns, performance and well-being, Publications Office
of the European Union, Luxembourg.
European Commission (2009) Industrial relations in Europe 2008, Luxembourg:
Publication Office of European Union.
Fields G.S. (2003a). “Accounting for income inequality and its change: A new
method with application to US earnings inequality”. Research in Labor Economics,
Vol 22.
Fields G.S. (2003b). “Regression-Based Decompositions: a new tool for managerial
decision-making”. Cornell University.
Fiorio, C.V. and Jenkins, S. P. (2003). “Regression-based inequality decomposition,
following Fields”. UKSUG.
40
Goux, D. y E. Maurin (1999): “Persistence of Interindustry Wage Differentials: A
Reexamination Using Matched Worker-Firm panel Data”, Journal of Labor
Economics, vol.17, n.º 3, págs. 492-533.
Greene, W.H. (1997): Econometric Analysis, 3rd edition. Ed.McMillan
Hausman, J.A. (1978): “Specification Tests in Econometrics”. Econometrica, 46(6).
Heshmati, A. (2004): “A Review of Decomposition of Income Inequality”. MTT
Economic Research and IZA Bonn. Discussion paper n.1221.
Hsiao, C. (1985): “Benefits and Limitations of Panel Data”. Econometrics Review, 4(1).
Katz, L.F.; Autor, D.H. (1999): “Change in the Wage Structure and Earnings
Inequality”, en O. Ashenfelter y D. Card (ed.) Handbook of Labor Economics, ed.
North-Holland.
Laroche M, Méretee M, Ruggeri G.C. “On the concept and dimensions of human
capital in a knowledge-based economy context”.98-01.
Lemieux T, Bentley MacLeod W, Parent D (2007): “Performance pay and wage
inequality”. IZA DPnº2850
Manna R. and Regoli. A. (2012). “Regression-based approaches for the
decomposition of income inequality in Italy, 1998-2008”. Rivista di statistica ufficiale
n1. Istituto nazionale di statistica 5.
Marsden, D (project coordinator) (1999). “Pay inequalities and economic
performance”. Project Report. Centre for Economic Performance. London
School of Economics.
Oaxaca, R. L. (1973). Male-Female Wage Differentials in Urban Labor Markets.
International Economic Review, 14(3), 693-709.
Pendleton A. ,Whitfield K., Bryson A.(2009). “The changing use of contingent pay
at the modern British workplace”. The Evolution of the Modern Workplace.
Cambridge University Press.
Pérez Trujillo M., Ruesga S. Da Silva Bichara J., Pérez Ortiz L. (2009).
“Competencia sindical en la negociación colectiva y efectos sobre el incremento
salarial pactado en convenio. El caso de España”. Comunicaciones Universidad
Autónoma de Madrid.
Palacio J.I, Simón H. (2002). “Dispersión salarial y negociación colectiva en
España”. Cuadernos Relaciones Laborales. Vol. 20 Núm 1 (2002): 169-187.
41
Palacio JIgnacio, Simón H. (2004). “Dispersión salarial entre establecimientos y
desigualdad salarial”. Revista de Economía Aplicada. Vol XII, núm 36, 2004 pp 47-
81. Universidad de Zaragoza.
Ramos R., Sanromá E., Simón H. (2014). “Public-Private sector wage Differentials
by type of contract: evidence from Spain”. Document de Treball XREAP2014-08
Reder, M. (1962): “Las diferencias interindustriales de salarios”, en A. Meixide (comp.)
El mercado de trabajo y la estructura salarial, Ministerio de Trabajo y Seguridad Social,
1988.
Shorrocks, A. F. (1982). Inequality Decomposition by Factor Components.
Econometrica, 50(1), 193-211
Shorrocks A. F. (1984), “Inequality Decomposition by Population Subgroups”,
Econometrica, 52, pp. 1369-85.
Simón H. (2009). “La desigualdad salarial en España: una perspectiva internacional y
temporal”. Investigaciones Económicas. Vol. XXXIII (3). 439-471.
Simón H. (ESES2002). “International Differences in Wage Inequality: a new glance
with European Matched Employer-Employee Data”.
Visser, J (2013). “Data Base on Institutional Characteristics of Trade Unions, Wage
Setting, State Intervention and Social Pacts, 1960-2011 (ICTWSS)”. Amsterdam
Institute for Advanced Labour Studies AIAS University of Amsterdam.
Wooldridge (2002a). Introductory Econometrics. A Modern Approach 2n.ed.